Closed ryanESX closed 1 year ago
I couldn't solve this besides stopping the bag early. I think the IMU readings saturate at that point.
On Sun, Jan 22, 2023 at 12:25 PM ryanESX @.***> wrote:
I tried to test indoor_forward_3 uzh fpv dataset. I found this dataset has hard landing in the end of bag file cause odometry divergence. Does it has anyway to resolve this problem?
— Reply to this email directly, view it on GitHub https://github.com/rpng/open_vins/issues/302, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAQ6TYRM62OIIVI4ZZJRJB3WTVUQZANCNFSM6AAAAAAUDC6QWU . You are receiving this because you are subscribed to this thread.Message ID: @.***>
The saturation of IMU. Do you mean it exceed the maximum of acceleration in IMU? For example, I set the max acceleration of the IMU in +-4g. But the hard landing make it 10g or any number exceed +-4g.
Yes, or that the higher accelerations are less sensitive. I have not looked into this problem / dataset directly, so take my thoughts with a grain of salt.
On Sun, Jan 22, 2023 at 12:55 PM ryanESX @.***> wrote:
The saturation of IMU. Do you mean it exceed the maximum of acceleration in IMU? For example, I set the max acceleration of the IMU in +-4g. But the hard landing make it 10g or any number exceed +-4g.
— Reply to this email directly, view it on GitHub https://github.com/rpng/open_vins/issues/302#issuecomment-1399561466, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAQ6TYXV5NEE4U7L5UQT543WTVYCHANCNFSM6AAAAAAUDC6QWU . You are receiving this because you commented.Message ID: @.***>
Okay, I got it. I will keep looking another datasets. I noticed uzh fpv dataset use Qualcomm Snapdrgon Flight. Their on-board IMU is commercial level. Euroc/RPGN dataset has better IMU(ADIS16448). Maybe it is another possible factor.
I tried to test indoor_forward_3 uzh fpv dataset. I found this dataset has hard landing in the end of bag file cause odometry divergence. Does it has anyway to resolve this problem?